Beginner · Module 04
Your Input Shapes the Output
After this, you will have seen that a short message produces a short response and a detailed message produces a detailed response — and you will know how to use that deliberately.
Intro
You've probably noticed by now that the same basic request can produce a thorough, usable response one day and something almost uselessly generic the next. That inconsistency usually isn't random — the AI mirrors the level of detail you gave it. This module shows you the mechanism, and gives you a template you can use from here on for any document-based task.
If you've sent a message to an AI and got back something too long, too shallow, or just not quite right, this module explains why. In your first month, you'll be doing a lot of tasks you haven't done before. This module shows you how to get better output faster, without trial and error.
The AI is not deciding how much to say at random. It is taking its cue from you. This module shows you how that works, and how to use it on purpose.
What is happening
When you send a short message, the AI reads it as: "a quick question, probably wants a quick answer." When you send a detailed message — one that explains your situation, what you need, and what a good response looks like — the AI reads it as: "this person wants something specific and thorough."
This is not a rule the AI follows. It is simply mirroring the style of what you sent. A one-sentence question gets a one-paragraph answer. A well-structured paragraph gets a well-structured response.
The good news is that you control this. If your responses are coming back thin or vague, more detail in your message will fix it. If they are coming back too long, keep your message shorter and add a stopping condition (B-06 covers this).
The exercise
This experiment takes about five minutes. You are going to send two versions of the same request and compare what comes back.
Step 1: Send Version A (short)
Copy this message, then personalise the topic in square brackets before sending:
Summarise this for me: [paste a short piece of text — a paragraph from an email, an article, anything you have nearby]
Send it. Read the response.
Step 2: Send Version B (detailed)
Start a new conversation so the AI is not building on the first one. Then copy this message and personalise every part in square brackets:
I'm a [your role] preparing for [brief description of what this is for — e.g. a team meeting, a presentation, a conversation with a client]. Summarise this in [format — e.g. 3 bullet points / a short paragraph] focusing on [the angle that matters most to you — e.g. the key risks / the next steps / the main finding]. Keep it under [length — e.g. 100 words / 5 bullet points].
[paste the same piece of text you used in Version A]
Send it. Read the response.
Step 3: Compare
Look at both responses side by side. Notice:
- Which one is longer?
- Which one is more specific to your situation?
- Which one would you actually use?
The difference is not because you used a magic formula. It is because you gave the AI more to work with.
The scenario: You've just come out of a client steering committee. There are meeting notes in your inbox. You need a summary for a status report that goes to a stakeholder who wasn't in the room.
Step 1: Send Version A — the message most people actually send
Copy this and send it with your meeting notes pasted at the end:
Summarise these meeting notes for me: [paste your meeting notes here]
Send it. Read what comes back.
Step 2: Start a new conversation and send Version B
Open a new conversation. Copy this message, fill in the brackets, and paste the same notes at the end:
I'm a senior account manager preparing the monthly status report for [client name]. Summarise these meeting notes in 4 bullet points covering decisions made, open issues, and agreed next steps. Write for a [Partner / Director / CFO — pick the one that fits] who wasn't in the meeting. Keep each bullet under 25 words. Don't include background context already covered in previous reports.
[paste the same meeting notes here]
Send it. Read what comes back.
Step 3: Compare
Which one could go directly into your report? Which one would you spend ten minutes editing? The difference is not that one used better technology. It is that one gave the AI a clear picture of what the output was for.
COPY — PERSONALISE — USE
The Version B starter is your template for this module. Here it is, clean:
I'm a [your role] preparing for [context]. Summarise this in [format] focusing on [angle]. Keep it under [length].
[your text here]
How to personalise it:
[your role] — what you do, in a few words. "Marketing assistant," "secondary school teacher," "small business owner."
[context] — one sentence on what this summary is for. "A catch-up call with my manager." "My weekly team update." "Preparing for a job interview."
[format] — how you want the information shaped. "3 bullet points." "A short paragraph." "5 numbered steps."
[angle] — what matters most from this text for your purpose. "The key decisions made." "What I need to do next." "Any risks mentioned."
[length] — how much you want. "Under 80 words." "No more than 5 bullet points." "One sentence per point."
I'm a [your role] preparing [what this output is for]. Summarise these [notes / document] in [format] covering [the specific angle]. Write for [intended reader]. Keep it under [length]. Don't include [what to leave out].
[paste your document here]
[your role] — "Senior account manager," "project lead," "engagement director."
[what this output is for] — "The monthly status report," "a client briefing ahead of next week's review."
[format] — "4 bullet points," "a short paragraph," "an executive summary under 150 words."
[the specific angle] — "Decisions made and next steps," "risks and open issues," "commercial implications."
[intended reader] — who will read this. The AI will adjust the level of assumed knowledge accordingly.
[length] — "Under 150 words," "no more than 5 bullets," "one sentence per point."
[what to leave out] — "Background context already covered," "actions not yet confirmed," "any caveats."
What good looks like
Version B should come back noticeably more tailored to your situation than Version A. It may not be perfect, but it should feel like it was written for you, not for anyone who asked the same generic question.
If this did not work
If both responses looked similar, check whether you personalised the Version B brackets fully. "I'm a [your role]" left unchanged will not help the AI; "I'm a part-time bookkeeper" will. Go back through each bracket and make sure it contains real, specific information.
If Version B came back very long despite the length you specified, that is worth noting — B-06 teaches you how to set a stopping condition that the AI is more likely to follow.
Next
Now that you have seen how your input style shapes the response, B-05 covers what happens when you ask for more than one thing at once — and why doing things one at a time works better.